How much do you care about prime (or other) stability?

[Gautam]

see poll

[eva2000]

> Just needs to be stable for the task at hand, whatever that may be

[Brian y.]

> Just needs to be stable for the task at hand, whatever that may be

Could not agree with you more eva

[chuchnit]

> Just needs to be stable for the task at hand, whatever that may be

Agreed I've got to the point to where I really only care if it gets to windows and makes a couple of runs. Other than that I just leave it stock for day to day use to save my time for hours of stability testing. I f I had more time maybe, but I've almost forgot my stable settings.

[ThugsRook]

there are 2 types of OCers, the drag racers and the daily drivers. obviously the drag racers just wanna make it to the finish line before they run out of LN, but the daily drivers are in it for the long haul....10 hours Prime

[Hyperhorn]

Depends on the task, Super Pi quick tests are the way to go.
btw I think those >10h test are pretty much useless. 2h and reducing the clocks a little bit afterwards to get a decent buffer is IMO the better way. Should not hurt to renounce 5 MHz FSB or so.

[Viss]

System is too stable/not pushed enough if it can do more then the bench you are trying to complete

[ThugsRook]

Depends on the task, Super Pi quick tests are the way to go.
btw I think those >10h test are pretty much useless. 2h and reducing the clocks a little bit afterwards to get a decent buffer is IMO the better way. Should not hurt to renounce 5 MHz FSB or so.

yup, even 10 hours can be completely worthless, Prime doesnt test everything. even if it wasnt worthless, its not a good idea to run redline 24/7. thats just asking for trouble.

but i think the OPs point is stability testing as a whole, not what you concider stable stability testing.

[zsamz_]

priming over an hour is useless to me
some wanna prime for 8hours thats up to them but i think ita a waste of time n puttin stress on hardware for nothin
2 passes of memtest n 1 hour or prime it plenty imho

[PcCI2iminal]

> Just needs to be stable for the task at hand, whatever that may be

+1

[philbrown23]

i personally think that people that prime their cpu for 10+ hours and wonder why it's degrading or dieing are idiots, you very rarely put your cpu through 100% load for 10+ hours straight so wtf is the point.

[sirheck]

If it will do what I want it to do, is all i want.

[SiGfever]

Testing for long hours with programs such as Linpak in my opinion do more harm than good. Most of us will never have stress requirements of our system that some of these programs produce. Stability testing is needed but an hour or two of Prime 95 "Blend" usually is sufficient for my system.

I do one hour Prime 95 "Blend" and then run WCG "BOINC" over night with all cores at 100%. If it is running in the morning, that is good enough for my needs.

[CyberDruid]

System is too stable/not pushed enough if it can do more then the bench you are trying to complete

I tend to agree.

I know my rig is not stable when it no longer boots

[Hyperhorn]

but i think the OPs point is stability testing as a whole, not what you concider stable stability testing.

Sure, but I think it's connected. Imagine there would be a 3 minute stability test, who works for 99,9%. Who would not use it?
So this thread is about how much effort you put in a stable, overclocked system.

[rge]

People are going to stress test to suit their needs...though some seem to think everyones needs should be the same.

Benching - not relevant. For gaming, surfing, etc - depends on persons past experiences with crashes or not versus their testing habits, or just how tolerant to crashes.

Then of course us "idiots" who crunch numbers, run multipass linear regression with multiple variables with 30,000 lines of data where cpu speed, accuracy, and stability is essential. I prefer prime stable 15+hrs settings that are as fast as possible, so no crashes with my data and does not take forever to crunch....and yeah, my cpu is never loaded like prime for 10 hours
Priming gives my computer a rest.
Different strokes...

[ThugsRook]

Sure, but I think it's connected. Imagine there would be a 3 minute stability test, who works for 99,9%. Who would not use it?
So this thread is about how much effort you put in a stable, overclocked system.

i agree ~ 1-2 hours per test is plenty

[CryptiK]

I use my pc for work, graphics rendering and also musical composition, so it must be stable to not risk losing work & time. I do appreciate a faster than stock system, so I overclock reasonably on safe voltages. I stability test with ~12 hours prime, memtest, linpack, whatever I can throw at it to make sure the whole system is stable. I'd rather wait a few hours for tests to complete (usually run overnight anyway) rather than spend a few hours re-doing work I lost due to an unexpected crash. I have had a pc that was 2 hours orthos stable crash while rendering a complex track I had written, so I use multiple apps to stability test, and test for as long as I can afford the time to, which is usually overnight so it doesn't impede my use of the pc.

However when I'm benching, I only care that it completes the task at hand and I can get the SS.

[The Nemesis]

I have two types of overclock goals. One is a 24/7 max system overclock which must be 8-12hrs prime stable. Second is Benchmark stable, benchmark being 3DMark 01-Vantage.

[chinaguy]

If I am working in Photoshop or doing Video work I dont want my PC crashing after 3 or 4 hours work! But sometimes its good to let the CPU run at top speed just to see what it can do.

I use to have a MG Metro Turbo with a turbo gate that could be set for .3 to .7 bar of pressure. Most of the time it was set to .3, but on the odd day I would set it to .7 just to burm some petrol for pleasure. You soon learnt about OVERHEATING and the need for a BIGGER radiator it that piece of junk! A friend had the same car but he had the .9 pressure gate set for 24/7 driving. We had to fit 2 cylinder heads as he blow them up at that pressure!

Its the same story with PCs and OC!

Boys and their Toys will never change its all the same, Cars, PC and Girls. Some boys like them the same way, HOT, FAST and DANGEROUS!

[dengyong]

5 passes of intel burn and 1 hour of occt is plenty for me.

[Fungus]

8 hours+ prime/burntest/occt stable for me. At least for CPU OC. Now if we only had a similar tools for Video Cards that really test them hard.

[dengyong]

8 hours+ prime/burntest/occt stable for me. At least for CPU OC. Now if we only had a similar tools for Video Cards that really test them hard.

OCCT v2.1.0.b06 has a very tough gpu test found in the occt folder. It really gets your card hot.

[carepolice]

^Furmark is great programm to show max GPU temps too! Usually it heats around +10c more than game usage.

And about prime stability, few hours then game to see real stability

[Movieman]

"Just needs to be stable for the task at hand, whatever that may be "
I don't run any of these apps, to me a waste of my time.
I build the systems using known proven methods, then I put them to work.
I come back and check them in an hour and then the next day.
Sort of the KISS method and honestly, it's never failed me.
I have 5 machines here that run at 100% load 24/7 for the past 2 years and in the work I do not one has ever tossed a bad WU at the program.